Browse Source

More efficient scalar multiplication for Short Weierstrass curves (#33)

* When a group element is a constant, precompute multiples of powers of two, and perform simple conditional additions (no doubling!).
* For short weierstrass curves, addition with a constant now uses mixed addition, which results in lower constraint weight.
* For short weierstrass curves, scalar multiplication now uses mixed addition, saving 1 constraint per bit of the scalar, along with lower constraint weight (at the cost of a small constant number of constraints to check for edge cases)
master
Pratyush Mishra 3 years ago
committed by GitHub
parent
commit
0162ef18bc
No known key found for this signature in database GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 211 additions and 55 deletions
  1. +159
    -37
      src/groups/curves/short_weierstrass/mod.rs
  2. +52
    -18
      src/groups/mod.rs

+ 159
- 37
src/groups/curves/short_weierstrass/mod.rs

@ -221,6 +221,46 @@ where
Ok(Self::new(x, y, z)) Ok(Self::new(x, y, z))
} }
/// Mixed addition, which is useful when `other = (x2, y2)` is known to have z = 1.
#[tracing::instrument(target = "r1cs", skip(self, x2, y2))]
pub(crate) fn add_mixed(&self, (x2, y2): (&F, &F)) -> Result<Self, SynthesisError> {
// Complete mixed addition formula from Renes-Costello-Batina 2015
// Algorithm 2
// (https://eprint.iacr.org/2015/1060).
// Below, comments at the end of a line denote the corresponding
// step(s) of the algorithm
//
// Adapted from code in
// https://github.com/RustCrypto/elliptic-curves/blob/master/p256/src/arithmetic/projective.rs
let three_b = P::COEFF_B.double() + &P::COEFF_B;
let (x1, y1, z1) = (&self.x, &self.y, &self.z);
let xx = x1 * x2; // 1
let yy = y1 * y2; // 2
let xy_pairs = ((x1 + y1) * &(x2 + y2)) - (&xx + &yy); // 4, 5, 6, 7, 8
let xz_pairs = (x2 * z1) + x1; // 8, 9
let yz_pairs = (y2 * z1) + y1; // 10, 11
let axz = mul_by_coeff_a::<P, F>(&xz_pairs); // 12
let bz3_part = &axz + z1 * three_b; // 13, 14
let yy_m_bz3 = &yy - &bz3_part; // 15
let yy_p_bz3 = &yy + &bz3_part; // 16
let azz = mul_by_coeff_a::<P, F>(z1); // 20
let xx3_p_azz = xx.double().unwrap() + &xx + &azz; // 18, 19, 22
let bxz3 = &xz_pairs * three_b; // 21
let b3_xz_pairs = mul_by_coeff_a::<P, F>(&(&xx - &azz)) + &bxz3; // 23, 24, 25
let x = (&yy_m_bz3 * &xy_pairs) - &yz_pairs * &b3_xz_pairs; // 28,29, 30
let y = (&yy_p_bz3 * &yy_m_bz3) + &xx3_p_azz * b3_xz_pairs; // 17, 26, 27
let z = (&yy_p_bz3 * &yz_pairs) + xy_pairs * xx3_p_azz; // 31, 32, 33
Ok(ProjectiveVar::new(x, y, z))
}
} }
impl<P, F> CurveVar<SWProjective<P>, <P::BaseField as Field>::BasePrimeField> impl<P, F> CurveVar<SWProjective<P>, <P::BaseField as Field>::BasePrimeField>
@ -288,18 +328,19 @@ where
// formulae are incomplete for even-order points. // formulae are incomplete for even-order points.
#[tracing::instrument(target = "r1cs")] #[tracing::instrument(target = "r1cs")]
fn enforce_prime_order(&self) -> Result<(), SynthesisError> { fn enforce_prime_order(&self) -> Result<(), SynthesisError> {
let r_minus_1 = (-P::ScalarField::one()).into_repr();
let mut result = Self::zero();
for b in BitIteratorBE::without_leading_zeros(r_minus_1) {
result.double_in_place()?;
if b {
result += self;
}
}
self.negate()?.enforce_equal(&result)?;
Ok(())
unimplemented!("cannot enforce prime order");
// let r_minus_1 = (-P::ScalarField::one()).into_repr();
// let mut result = Self::zero();
// for b in BitIteratorBE::without_leading_zeros(r_minus_1) {
// result.double_in_place()?;
// if b {
// result += self;
// }
// }
// self.negate()?.enforce_equal(&result)?;
// Ok(())
} }
#[inline] #[inline]
@ -308,9 +349,11 @@ where
// Complete doubling formula from Renes-Costello-Batina 2015 // Complete doubling formula from Renes-Costello-Batina 2015
// Algorithm 3 // Algorithm 3
// (https://eprint.iacr.org/2015/1060). // (https://eprint.iacr.org/2015/1060).
// Below, comments at the end of a line denote the corresponding
// step(s) of the algorithm
// //
// Adapted from code in // Adapted from code in
// https://github.com/RustCrypto/elliptic-curves/blob/master/p256/src/arithmetic.rs
// https://github.com/RustCrypto/elliptic-curves/blob/master/p256/src/arithmetic/projective.rs
let three_b = P::COEFF_B.double() + &P::COEFF_B; let three_b = P::COEFF_B.double() + &P::COEFF_B;
let xx = self.x.square()?; // 1 let xx = self.x.square()?; // 1
@ -346,6 +389,57 @@ where
fn negate(&self) -> Result<Self, SynthesisError> { fn negate(&self) -> Result<Self, SynthesisError> {
Ok(Self::new(self.x.clone(), self.y.negate()?, self.z.clone())) Ok(Self::new(self.x.clone(), self.y.negate()?, self.z.clone()))
} }
/// Computes `bits * self`, where `bits` is a little-endian
/// `Boolean` representation of a scalar.
#[tracing::instrument(target = "r1cs", skip(bits))]
fn scalar_mul_le<'a>(
&self,
bits: impl Iterator<Item = &'a Boolean<<P::BaseField as Field>::BasePrimeField>>,
) -> Result<Self, SynthesisError> {
if self.is_constant() {
// Compute 2^i * self for i in 0..bits.len()
// (these will be used by`precomputed_base_scalar_mul_le`, to perform
// a conditional addition.).
//
// TODO: if `bits.len()` is small n, it might be cheaper to
// conditionally select between 2^n options.
let mut value = self.value()?;
let bits_and_multiples = bits
.map(|b| {
let multiple = value;
value.double_in_place();
(b, multiple)
})
.collect::<ark_std::vec::Vec<_>>();
let mut result = self.clone();
result.precomputed_base_scalar_mul_le(
bits_and_multiples.iter().map(|&(ref b, ref c)| (*b, c)),
)?;
Ok(result)
} else {
// Computes the standard big-endian double-and-add algorithm
// (Algorithm 3.27, Guide to Elliptic Curve Cryptography)
// the input is little-endian, so we need to reverse it to get it in
// big-endian.
let mut bits = bits.collect::<Vec<_>>();
bits.reverse();
let mut res = Self::zero();
let self_affine = self.to_affine()?;
for bit in bits {
res.double_in_place()?;
let tmp = res.add_mixed((&self_affine.x, &self_affine.y))?;
res = bit.select(&tmp, &res)?;
}
// The foregoing algorithm relies on mixed addition, and so does not
// work when the input (`self`) is zero. We hence have to perform
// a check to ensure that if the input is zero, then so is the output.
// The cost of this check should be less than the benefit of using
// mixed addition in almost all cases.
self_affine.infinity.select(&Self::zero(), &res)
}
}
} }
impl<P, F> ToConstraintFieldGadget<<P::BaseField as Field>::BasePrimeField> for ProjectiveVar<P, F> impl<P, F> ToConstraintFieldGadget<<P::BaseField as Field>::BasePrimeField> for ProjectiveVar<P, F>
@ -385,41 +479,69 @@ impl_bounded_ops!(
add, add,
AddAssign, AddAssign,
add_assign, add_assign,
|this: &'a ProjectiveVar<P, F>, other: &'a ProjectiveVar<P, F>| {
// Complete addition formula from Renes-Costello-Batina 2015
// Algorithm 1
|mut this: &'a ProjectiveVar<P, F>, mut other: &'a ProjectiveVar<P, F>| {
// Implement complete addition for Short Weierstrass curves, following
// the complete addition formula from Renes-Costello-Batina 2015
// (https://eprint.iacr.org/2015/1060). // (https://eprint.iacr.org/2015/1060).
// //
// Adapted from code in
// https://github.com/RustCrypto/elliptic-curves/blob/master/p256/src/arithmetic.rs
// We special case handling of constants to get better constraint weight.
if this.is_constant() {
// we'll just act like `other` is constant.
core::mem::swap(&mut this, &mut other);
}
let three_b = P::COEFF_B.double() + &P::COEFF_B;
if other.is_constant() {
// The value should exist because `other` is a constant.
let other = other.value().unwrap();
if other.is_zero() {
// this + 0 = this
this.clone()
} else {
// We'll use mixed addition to add non-zero constants.
let x = F::constant(other.x);
let y = F::constant(other.y);
this.add_mixed((&x, &y)).unwrap()
}
} else {
// Complete addition formula from Renes-Costello-Batina 2015
// Algorithm 1
// (https://eprint.iacr.org/2015/1060).
// Below, comments at the end of a line denote the corresponding
// step(s) of the algorithm
//
// Adapted from code in
// https://github.com/RustCrypto/elliptic-curves/blob/master/p256/src/arithmetic/projective.rs
let three_b = P::COEFF_B.double() + &P::COEFF_B;
let (x1, y1, z1) = (&this.x, &this.y, &this.z);
let (x2, y2, z2) = (&other.x, &other.y, &other.z);
let xx = &this.x * &other.x; // 1
let yy = &this.y * &other.y; // 2
let zz = &this.z * &other.z; // 3
let xy_pairs = ((&this.x + &this.y) * &(&other.x + &other.y)) - (&xx + &yy); // 4, 5, 6, 7, 8
let xz_pairs = ((&this.x + &this.z) * &(&other.x + &other.z)) - (&xx + &zz); // 9, 10, 11, 12, 13
let yz_pairs = ((&this.y + &this.z) * &(&other.y + &other.z)) - (&yy + &zz); // 14, 15, 16, 17, 18
let xx = x1 * x2; // 1
let yy = y1 * y2; // 2
let zz = z1 * z2; // 3
let xy_pairs = ((x1 + y1) * &(x2 + y2)) - (&xx + &yy); // 4, 5, 6, 7, 8
let xz_pairs = ((x1 + z1) * &(x2 + z2)) - (&xx + &zz); // 9, 10, 11, 12, 13
let yz_pairs = ((y1 + z1) * &(y2 + z2)) - (&yy + &zz); // 14, 15, 16, 17, 18
let axz = mul_by_coeff_a::<P, F>(&xz_pairs); // 19
let axz = mul_by_coeff_a::<P, F>(&xz_pairs); // 19
let bzz3_part = &axz + &zz * three_b; // 20, 21
let bzz3_part = &axz + &zz * three_b; // 20, 21
let yy_m_bzz3 = &yy - &bzz3_part; // 22
let yy_p_bzz3 = &yy + &bzz3_part; // 23
let yy_m_bzz3 = &yy - &bzz3_part; // 22
let yy_p_bzz3 = &yy + &bzz3_part; // 23
let azz = mul_by_coeff_a::<P, F>(&zz);
let xx3_p_azz = xx.double().unwrap() + &xx + &azz; // 25, 26, 27, 29
let azz = mul_by_coeff_a::<P, F>(&zz);
let xx3_p_azz = xx.double().unwrap() + &xx + &azz; // 25, 26, 27, 29
let bxz3 = &xz_pairs * three_b; // 28
let b3_xz_pairs = mul_by_coeff_a::<P, F>(&(&xx - &azz)) + &bxz3; // 30, 31, 32
let bxz3 = &xz_pairs * three_b; // 28
let b3_xz_pairs = mul_by_coeff_a::<P, F>(&(&xx - &azz)) + &bxz3; // 30, 31, 32
let x = (&yy_m_bzz3 * &xy_pairs) - &yz_pairs * &b3_xz_pairs; // 35, 39, 40
let y = (&yy_p_bzz3 * &yy_m_bzz3) + &xx3_p_azz * b3_xz_pairs; // 24, 36, 37, 38
let z = (&yy_p_bzz3 * &yz_pairs) + xy_pairs * xx3_p_azz; // 41, 42, 43
let x = (&yy_m_bzz3 * &xy_pairs) - &yz_pairs * &b3_xz_pairs; // 35, 39, 40
let y = (&yy_p_bzz3 * &yy_m_bzz3) + &xx3_p_azz * b3_xz_pairs; // 24, 36, 37, 38
let z = (&yy_p_bzz3 * &yz_pairs) + xy_pairs * xx3_p_azz; // 41, 42, 43
ProjectiveVar::new(x, y, z)
}
ProjectiveVar::new(x, y, z)
}, },
|this: &'a ProjectiveVar<P, F>, other: SWProjective<P>| { |this: &'a ProjectiveVar<P, F>, other: SWProjective<P>| {
this + ProjectiveVar::constant(other) this + ProjectiveVar::constant(other)

+ 52
- 18
src/groups/mod.rs

@ -92,39 +92,73 @@ pub trait CurveVar:
&self, &self,
bits: impl Iterator<Item = &'a Boolean<ConstraintF>>, bits: impl Iterator<Item = &'a Boolean<ConstraintF>>,
) -> Result<Self, SynthesisError> { ) -> Result<Self, SynthesisError> {
let mut res = Self::zero();
let mut multiple = self.clone();
for bit in bits {
let tmp = res.clone() + &multiple;
res = bit.select(&tmp, &res)?;
multiple.double_in_place()?;
if self.is_constant() {
// Compute 2^i * self for i in 0..bits.len()
// (these will be used by`precomputed_base_scalar_mul_le`, to perform
// a conditional addition.).
//
// TODO: if `bits.len()` is small n, it might be cheaper to
// conditinally select between 2^n options.
let mut value = self.value().unwrap();
let bits_and_multiples = bits
.map(|b| {
let multiple = value;
value.double_in_place();
(b, multiple)
})
.collect::<ark_std::vec::Vec<_>>();
let mut result = self.clone();
result.precomputed_base_scalar_mul_le(
bits_and_multiples.iter().map(|&(ref b, ref c)| (*b, c)),
)?;
Ok(result)
} else {
// Computes the standard little-endian double-and-add algorithm
// (Algorithm 3.27, Guide to Elliptic Curve Cryptography)
let mut res = Self::zero();
let mut multiple = self.clone();
for bit in bits {
let tmp = res.clone() + &multiple;
res = bit.select(&tmp, &res)?;
multiple.double_in_place()?;
}
Ok(res)
} }
Ok(res)
} }
/// Computes a `I * self` in place, where `I` is a `Boolean` *little-endian* /// Computes a `I * self` in place, where `I` is a `Boolean` *little-endian*
/// representation of the scalar. /// representation of the scalar.
/// ///
/// The base powers are precomputed power-of-two multiples of a single
/// The bases are precomputed power-of-two multiples of a single
/// base. /// base.
#[tracing::instrument(target = "r1cs", skip(scalar_bits_with_base_powers))]
#[tracing::instrument(target = "r1cs", skip(scalar_bits_with_bases))]
fn precomputed_base_scalar_mul_le<'a, I, B>( fn precomputed_base_scalar_mul_le<'a, I, B>(
&mut self, &mut self,
scalar_bits_with_base_powers: I,
scalar_bits_with_bases: I,
) -> Result<(), SynthesisError> ) -> Result<(), SynthesisError>
where where
I: Iterator<Item = (B, &'a C)>, I: Iterator<Item = (B, &'a C)>,
B: Borrow<Boolean<ConstraintF>>, B: Borrow<Boolean<ConstraintF>>,
C: 'a, C: 'a,
{ {
for (bit, base_power) in scalar_bits_with_base_powers {
let new_encoded = self.clone() + *base_power;
*self = bit.borrow().select(&new_encoded, self)?;
// Computes the standard little-endian double-and-add algorithm
// (Algorithm 3.26, Guide to Elliptic Curve Cryptography)
// Let `original` be the initial value of `self`.
let mut result = Self::zero();
for (bit, base) in scalar_bits_with_bases {
// Compute `self + 2^i * original`
let self_plus_base = result.clone() + *base;
// If `bit == 1`, set self = self + 2^i * original;
// else, set self = self;
result = bit.borrow().select(&self_plus_base, &result)?;
} }
*self = result;
Ok(()) Ok(())
} }
/// Computes a `\sum_j I_j * B_j`, where `I_j` is a `Boolean`
/// Computes `Σⱼ(scalarⱼ * baseⱼ)` for all j,
/// where `scalarⱼ` is a `Boolean` *little-endian*
/// representation of the j-th scalar. /// representation of the j-th scalar.
#[tracing::instrument(target = "r1cs", skip(bases, scalars))] #[tracing::instrument(target = "r1cs", skip(bases, scalars))]
fn precomputed_base_multiscalar_mul_le<'a, T, I, B>( fn precomputed_base_multiscalar_mul_le<'a, T, I, B>(
@ -137,11 +171,11 @@ pub trait CurveVar:
B: Borrow<[C]>, B: Borrow<[C]>,
{ {
let mut result = Self::zero(); let mut result = Self::zero();
// Compute ∏(h_i^{m_i}) for all i.
for (bits, base_powers) in scalars.zip(bases) {
let base_powers = base_powers.borrow();
// Compute Σᵢ(bitᵢ * baseᵢ) for all i.
for (bits, bases) in scalars.zip(bases) {
let bases = bases.borrow();
let bits = bits.to_bits_le()?; let bits = bits.to_bits_le()?;
result.precomputed_base_scalar_mul_le(bits.iter().zip(base_powers))?;
result.precomputed_base_scalar_mul_le(bits.iter().zip(bases))?;
} }
Ok(result) Ok(result)
} }

Loading…
Cancel
Save